gemini-gen

Pass

Audited by Gen Agent Trust Hub on Apr 11, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The scripts/gemini.mjs script implements a data flow where local files are read and transmitted to an external endpoint (renoise.ai). This represents a risk as the tool could be directed to read and exfiltrate sensitive files from the local filesystem.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes content from external files (images and videos) which could contain hidden instructions designed to manipulate the agent's behavior.\n
  • Ingestion points: Local file data is ingested through the fs.readFile call in scripts/gemini.mjs prompted by the --file command-line argument.\n
  • Boundary markers: The prompt construction lacks clear delimiters or protective instructions to distinguish between the system's analysis instructions and potentially malicious content embedded within the files.\n
  • Capability inventory: The skill possesses both file system read access and the ability to make outbound network requests via the native fetch API.\n
  • Sanitization: No validation or sanitization of the file content is performed before it is sent to the AI model.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 11, 2026, 01:34 AM
Security Audit — agent-trust-hub — gemini-gen